Warehouses Capture Most Stack Value
Earl Lee, co-founder and CEO of HeadsUp, on the modern data stack value chain
The warehouse wins first because every upstream and downstream tool turns into more warehouse spend. When a team lands raw data in Snowflake, Redshift, BigQuery, or Databricks, then transforms it with dbt, syncs it out with reverse ETL, or queries it in an app, each step burns compute and keeps the warehouse at the center of the workflow. That makes the warehouse the toll booth for the whole stack, even when another vendor owns the user interface.
-
This is why lightweight categories like reverse ETL can be useful without capturing the deepest economics. Their job is mostly to move already modeled data from the warehouse into Salesforce, HubSpot, or Braze. The customer may pay the app vendor a subscription, but the heaviest ongoing usage meter still often sits in the warehouse.
-
The second big value pool sits above the warehouse in semantic meaning. dbt became important by letting analytics engineers write business logic once, test it, version it, and turn raw tables into trusted definitions like revenue, activation, or CAC. That logic is harder to replace than a connector because it encodes how the business actually works.
-
That creates a push and pull in the market. Warehouses are rebundling more of the stack to capture more spend, while tools like dbt argue the business logic should live outside any single cloud because large companies use multiple warehouses and BI tools. The control point is shifting from just storing data to defining it consistently across systems.
Going forward, the biggest winners are likely to be the platforms that either meter core compute or own the shared definitions every downstream app depends on. More stack consolidation will happen around Snowflake, Databricks, and the hyperscalers, while semantic and control plane layers keep growing as the neutral place where companies define business logic once and reuse it everywhere.